Plasma reactors employed in deposition processing of semiconductor wafers (e.g., physical vapor deposition or chemical vapor deposition) tend to accumulate contaminants on the interior chamber surfaces. Such contamination occurs because the same material deposited on the wafer is also deposited on the interior chamber surfaces. Such deposits on the interior chamber surfaces, if allowed to accumulate significantly, can fall onto the semiconductor wafer and thus destroy devices formed thereon, or change the electrical properties of the chamber and thereby reduce control over the process. Process control in the wafer processing chamber is critical. Therefore, the chamber interior must be cleaned periodically to remove such contaminants.
A recently proposed technique for cleaning a chamber of a deposition reactor is to form a plasma in another (exterior) chamber of a highly reactive etch species, such as fluorine, and deliver the fluorine plasma to the interior of the wafer processing chamber. Since fluorine gas (F2) is highly toxic and difficult to handle as a process gas, it is not generally commercially available, so that another more stable gas such as NF3 is supplied as the process gas, and this gas is ionized in the external chamber to form the required fluorine plasma. NF3 is easy to handle in the gas supply because it is relatively stable, but this stability means that plasma activation requires large power, resulting in a harsh plasma environment. Use of the external chamber spares the wafer processing chamber from exposure limits the etch rate during the cleaning operation. In order to reduce recombination of free fluorine prior to reaching the wafer processing chamber interior walls, the secondary chamber is placed virtually on top of the wafer processing chamber, and a direct port-to-port path for the fluorine is provided between the two chambers. Thus, fluorine enters the wafer processing chamber from a single port, typically in the chamber ceiling. While this approach does provide a limited improvement in performance, the single entry point of the fluorine dictates a non-uniform distribution of fluorine in the chamber, so that chamber interior components nearer the ceiling are cleaned sooner, while chamber components nearer the floor of the chamber are cleaned last. This latter aspect contributes to the amount of time required to clean the chamber.
Attempts have been made to reduce the time required to clean the chamber (by increasing the etch rate during the chamber clean operation). This entailed increasing the amount of fluorine gas flow to the secondary chamber and increasing the plasma source power applied to the secondary chamber. The result has been an increase in the amount of free fluorine ions and radicals furnished to the wafer processing chamber from the secondary chamber. For example, if the secondary chamber is an inductively coupled plasma reactor, then the RF power applied to its inductive antenna is increased.
Surprisingly, such attempts have resulted in very little if any improvement in etch rate and the time required to clean the wafer processing chamber. Moreover, such an approach involves certain disadvantages. Specifically, the RF power generator, the impedance match circuit and the antenna of the secondary chamber must be capable of sustaining high RF power levels (e.g., thousands of Watts), which entails greater expense. Moreover, the increase in fluorine gas flow is limited by environmental restrictions on the use of fluorine gas.
The present inventors have discovered that the failure to realize a concomitant improvement in etch rate in the wafer processing chamber in response to greater supply of free fluorine from the secondary chamber arises from several factors. First, the single entry port provided for the incoming free fluorine ions and radicals faces the center of the chamber interior in order to realize a more even distribution of fluorine across the chamber walls. This however directs most of the free fluorine to the center of the chamber interiorxe2x80x94and away from the chamber wallsxe2x80x94so that much of the fluorine must travel (diffuse) from the center toward the chamber walls before it can attack the accumulated contamination on the walls. During this travel, much of the free fluorine recombines into F2, thereby reducing the etch rate.
Second, the introduction of the free fluorine at the ceiling propels the free fluorine toward the pumping annulus near the bottom of the chamber, so that much of the free fluorine is pumped out of the chamber before it ever diffuses toward the chamber side walls or other chamber components. In fact, we have measured the proportion of fluorine that fails to diffuse toward the chamber walls, and find that it is about half the free fluorine. Thus, the free fluorine is introduced over the center of the chamber and travels downwardly toward the chamber floor, where much of it is pumped out of the chamber before it can reach the chamber walls. Apparently, increasing the amount of free fluorine flowing into the wafer processing chamber merely increases the amount of fluorine pumped out through the pumping annulus without appreciably increasing the amount of free fluorine reaching the chamber side walls.
A further problem is that introduction of the free fluorine at the single entry port at the ceiling provides an uneven distribution of free fluorine within the chamber, so that chamber walls and components near the ceiling are cleaned first while chamber walls and components nearer the chamber floor are cleaned more slowly and are therefore the last to be cleaned. The cleaning step must be carried on for the amount of time required to clean the components and walls near the floorxe2x80x94the last components to be cleaned. Thus, the uneven distribution of free fluorine slows down the cleaning process.
These problems are exacerbated as wafer sizexe2x80x94and hence chamber sizexe2x80x94increases. The volume of a larger chamber increases with the square of the chamber radius, and it is this volume in which much of the free fluorine resides before it either recombines or is removed through the pumping annulus. Thus, as wafer size scales upwardly, the foregoing problems worsen.
One technique we considered in an attempt to overcome the foregoing problems is to re-ionize the F2 molecules into which the free fluorine had recombined. Such re-ionization could be carried out in the wafer processing chamber (xe2x80x9cin-situxe2x80x9d) using existing RF power applicators (e.g., an RF generator connected to the wafer support pedestal and/or an RF generator connected to an inductive antenna of the wafer processing chamber). The problem with this approach is that such re-ionization would be effective only near such RF power applicators, and would be relatively ineffective otherwise. That is, for chamber walls and components relatively far from any RF power applicator of the chamber, there would be less re-ionization (or none at all) and therefore the time to clean such distant components would be little improved, or not improved at all. Therefore, in-situ cleaning appears to provide no solution to the problem.
In summary, it has seemed impractical to completely fulfill the prior art goal of minimizing recombination in order to deliver free fluorine directly from the secondary chamber to the interior walls and components of the wafer processing chamber
An electro-negative cleaning or etchant gas, such as fluorine, that was ionized from a stable supply gas such as NH3 in a secondary chamber and recombined in the primary chamber, is re-ionized within the primary chamber by electron attachment by ionizing an electron donor gas, such as helium, in the primary chamber.